--Acted as technical lead for a team of 5 data scientists in developing a PySpark feature library and set of machine learning models for search result ranking, resulting in a significant increase in conversion.-- Created a framework for describing the customer lifecycle in daytime pet services (dog walking and daycare), and formalized metrics to facilitate timely marketplace health monitoring based on transition rates.-- Designed and implemented a survival-based model to optimize retention and lifetime value in sitter-client relationships.

--Mentored Data Science fellows through project advice and mock interviews.--Conducted interviews for program applicants.--Participated in alumni panels.

--Implemented an email personalization approach that improved clickthrough rates compared to previously used tools.--Collaborated with engineering to build a microservice that uses fuzzy matching to extract review snippets that mention sponsoring businesses. Also used a similar approach to link thousands of unattributed reviews to providers.--Built a site feature to collect data on semantically duplicative user-generated questions.

--Implemented a Python-based Support Vector Machine classification algorithm for identifying sponsored content in YouTube videos based on captions.--Pulled 1000+ video captions from YouTube’s API, stored them in a SQL database--Deployed algorithm through a Google Chrome extension and web app powered by Flask, hosted on Amazon Web Services.

In continuation of my work as a Graduate Research Assistant at Purdue University:--Refined R implementation of MCMC model estimation algorithm, improving efficiency >10-fold and expanding applicability to more general experimental designs.--Developed an R package (msiCompare).--Submitted 1st-authored manuscript to a peer-reviewed journal.

--Developed a Bayesian statistical modeling and testing framework for mass spectrometry imaging data, in order to identify differences in the composition of biological tissues.--Prototyped a Markov Chain Monte Carlo estimation algorithm for the model in R. Simulated >50 thousand images for model evaluation, concluding that the contributed model showed improved sensitivity and specificity over existing methods. --Co-maintained the R package Cardinal, winner of the 2015 John C. Chambers statistical software award from the American Statistical Association.--Presented talks at 2 national conferences (US HUPO 2015 & ASMS 2016) and contributed to 2 peer reviewed publications.

--Course: STAT 301: Elementary Statistical Methods.--Instructed lab sections using the SPSS statistical computing software.--Graded exams and lab assignments.--Held office hours.

--Directly mentored 14 underrepresented, female students preparing forgraduate programs in the mathematical sciences.--Facilitated daily homework problem sessions.--Organized community-building activities.

--Public Health Group, manager J. Kaufman.--Evaluated the performance of a statistical method for predicting the sources of food-borne illness epidemics, based on spatially-registered shopping data. --Contributed to 1 peer-reviewed publication.